DOBOT X-Trainer
- Dual-arm teleoperated platform with 0.05 mm precision Nova 2 cobots for accurate motion capture
- 25 Hz motion interface delivers 150% faster response than comparable platforms
- Cuts routine scenario training time by 70%, completing tasks like dishwashing in 2 hours
- Real-time interference correction during autonomous task execution
- Complete SDK with open APIs for joint angles, gripper control, and RGB/depth camera data
- Lightweight, foldable base with EVA foam padding for safe, fast deployment
- Trusted dealerDirect manufacturer partnership
- Pan-India deliveryInsured shipping, all states
- Post-sale supportTraining + service included
What makes it work
DOBOT X-Trainer — AI Data Collection & Training Robotic System
A dual-arm bimanual teleoperation platform built for AGI research, imitation learning, and real-world AI skill training. Industrial-grade precision meets rapid autonomous deployment.
Why the X-Trainer Stands Apart
70% Faster Skill Acquisition
Training time is dramatically reduced by 70%. Routine scenarios such as dish-washing or clothes-folding can be completed in as little as 2 hours, enabling fast iteration on AI tasks.
25 Hz High-Performance Interface
The end-to-end motion interface operates at 25 Hz, delivering a response speed 150% faster than comparable systems — providing stable, real-time control during teleoperation and inference.
±0.05 mm Repeat Positioning
Each slave arm achieves ±0.05 mm repeat positioning accuracy, meeting industrial-grade standards for sensitive data collection tasks and precision laboratory experiments.
625 mm → 1200 mm Workspace
A 625 mm per-arm reach expands to a combined 1200 mm workspace in dual-arm mode. With full 6-axis flexibility per arm, the X-Trainer covers a broad operational envelope for diverse data capture tasks.
5-Level Collision Detection
The integrated Nova 2 cobot features 5-level collision detection, ISO 15066 certification, and master hand drop detection — with automatic master-slave desynchronisation on power loss to prevent accidents.
Full Data Pipeline with Open Access
A complete SDK covers data collection, model training, and autonomous inference. Open APIs expose joint angles, gripper control, and camera RGB/depth data for seamless integration with custom AI models and neural networks.
Technical Specifications
| Specification | Details |
|---|---|
| Slave Arms (Nova 2 Cobots) | |
| Degrees of Freedom (per arm) | 6-Axis |
| Repeat Positioning Accuracy | ±0.05 mm |
| Reach (per arm) | 625 mm |
| Dual-Arm Workspace | 1200 mm (expandable in dual-arm mode) |
| Single-Arm Payload | 2 kg |
| Dual-Arm Payload | 3 kg |
| Maximum TCP Speed | 1.6 m/s |
| Collision Detection Levels | 5-Level |
| Safety Certification | ISO 15066 |
| Master Arms (Teleoperation) | |
| Total System Degrees of Freedom | 14 DoF (7 per master arm) |
| Master Hand Load | Under 400 g |
| Joint Construction | 3D Printed Monolithic Joints |
| Control Buttons | Unlock Button + Record Button (per master hand) |
| Actuation | Spring-Assisted Chassis, High-Performance Servos |
| Motion Interface & AI | |
| Motion Interface Frequency | 25 Hz (End-to-End) |
| Response Speed Improvement | 150% faster than comparable systems |
| Training Time Reduction | 70% shorter than conventional methods |
| Routine Task Training Duration | ~2 hours (e.g., dish-washing, clothes-folding) |
| Imitation Learning | Instant Imitation Learning Networks |
| Interference Correction | Real-Time Disturbance Correction (Autonomous) |
| Camera & Sensing | |
| Included Camera | Intel RealSense D405 (RGB + Depth) |
| API Data Access | Master/Slave Joint Angles, Gripper Control, Camera RGB/Depth |
| Safety & Protection | |
| Workspace Protection | Flexible EVA Foam Padding (End-Effector Drop Cushioning) |
| Safety Events | Auto Master-Slave De-sync on Power Loss / Alarms; Drop Detection |
| Low-Speed Sync | Low-Speed Master-Slave Synchronisation (Collision Prevention) |
| Software & Development | |
| SDK Coverage | Data Collection → Model Training → Autonomous Inference |
| API Type | Open APIs (Custom AI Model Integration) |
| Export Formats | Specialised Data Formats for Domain-Specific Neural Networks |
| Documentation | Full Pipeline Documentation Included |
A Unified Platform for Real-World AI Training
The DOBOT X-Trainer is an advanced AI data collection and training robotic system built to accelerate AGI research and bring practical AI projects to life. It pairs a dual-arm bimanual teleoperation framework with an instant imitation learning architecture — enabling researchers and engineers to teach complex manipulation tasks in a fraction of the time traditionally required.
Where conventional AI training pipelines demand days of data gathering and simulation, the X-Trainer compresses the cycle. Routine scenario training such as dish-washing or folding clothes can be completed autonomously in as little as 2 hours — a 70% reduction in training time. The system bridges the gap between human demonstration and autonomous robot execution with minimal friction.
Designed for the Human Operator
Weighing in at under 400g per master hand, the X-Trainer's teleoperation interface prioritises operator comfort during extended sessions. Monolithic 3D-printed joints reduce part count and mechanical play, while a spring-assisted chassis and high-performance servos ensure motion feels natural and fatigue-free.
Each master hand features a dedicated unlock button and a record button. The unlock function allows instant locking and unlocking of the arm's pose — letting operators rest without losing position — while the record button provides direct, in-hand control over data capture timing. The result is an ergonomic, intuitive workflow that makes quality data collection straightforward.
Industrial Accuracy, Research Flexibility
The X-Trainer's slave arms achieve ±0.05 mm repeat positioning accuracy — a standard drawn from industrial collaborative robotics and applied directly to the AI training context. Each arm offers a 625 mm reach that expands to 1200 mm in dual-arm operation, with full 6-axis articulation to cover wide and varied workspaces.
Payload capacity of 2 kg per arm (3 kg combined in dual-arm mode) and a top TCP speed of 1.6 m/s ensure the platform handles a broad spectrum of tasks, from precise handling of small objects to faster, broader manipulation sequences that generate rich, diverse training datasets.
Safety Engineered at Every Layer
The X-Trainer's Nova 2 cobot arms carry ISO 15066 certification and feature 5-level collision detection, enabling safe side-by-side operation with human researchers. On any power interruption or alarm event, the master-slave connection is automatically severed, preventing unintended arm movement.
Master hand drop detection and low-speed master-slave synchronisation add further protection against accidental contact. The entire workspace is lined with flexible EVA foam padding that cushions end-effector drops and reduces wear on grippers and arm structures — practical protection for a platform that will see frequent and intensive daily use.
Secondary Development & AI Integration
Instant Imitation Learning
Human demonstrations are captured and converted into reusable model inputs through the X-Trainer's imitation learning networks — compressing weeks of data collection into hours.
25 Hz Motion Interface
End-to-end latency of just 40 ms ensures teleoperation commands reach the slave arms with minimal delay, keeping motion smooth and data temporally consistent.
Open API Access
APIs expose joint angles (master and slave), gripper state, and Intel RealSense D405 RGB/depth streams — enabling plug-in integration with custom AI models and research pipelines.
Full SDK Included
The complete SDK covers the entire workflow: data collection, model training, and autonomous inference, with detailed documentation guiding users through each stage of the pipeline.
Specialised Data Export
Datasets can be exported in formats compatible with domain-specific neural networks, removing conversion overhead and letting teams move directly from collection to training.
Interference Resistance
During autonomous operation, the X-Trainer detects and responds to real-time disturbances — for example, recognising unexpected stains during a dish-washing task and correcting without human intervention.
Performance at a Glance
X-Trainer Key Performance Metrics
Frequently Asked Questions
The DOBOT X-Trainer is a dual-arm bimanual teleoperation and AI training system designed for researchers, university labs, and AI engineers working on AGI research, embodied intelligence projects, and practical AI robotics competitions. It provides a complete hardware and software environment for collecting training data, building imitation learning models, and deploying autonomous tasks — all in one integrated platform.
The X-Trainer's instant imitation learning network reduces training time by approximately 70% compared to conventional approaches. A routine task such as dish-washing or folding clothes can be taught and autonomously executed in as little as 2 hours. More complex tasks will require additional demonstration cycles, but the 25 Hz motion interface and real-time interference correction significantly accelerate the full pipeline from human demonstration to autonomous execution.
The full system bundle (model 8000000050) includes two Nova 2 collaborative robotic arms, two slave hand grippers, and an Intel RealSense D405 depth camera. The Nova 2 arms serve as the slave hands — executing the motions captured via the master teleoperation handles. Please confirm bundle contents with your authorised reseller at the time of purchase, as configurations may vary by region.
The X-Trainer ships with a complete SDK covering data collection, model training, and autonomous inference. Open APIs give developers access to master and slave joint angles, gripper control, and camera RGB/depth streams — enabling integration with custom AI models built on any major framework. The SDK also supports exporting data in specialised formats for direct use with domain-specific neural networks. Projects like LeRobot format conversion and Isaac Sim digital twin integration have been demonstrated on the platform through the official GitHub repository.
Yes. Safety is a core design priority. The integrated Nova 2 cobots carry ISO 15066 certification and 5-level collision detection for safe human-robot proximity. Automatic master-slave desynchronisation triggers on power loss or alarms, and low-speed master-slave sync prevents unintended collisions during setup. The entire workspace is padded with flexible EVA foam to cushion accidental drops of the end-effector, minimising damage risk to both the robot and surrounding objects. These features collectively make the X-Trainer well-suited to research lab and university classroom deployments.
The full sheet
Related products
-
·XBM·00
-
·XBM·01
-
·XBM·02
-
·XBM·03
There are no reviews yet.